During January 2026, WPC enhanced the jeejeelee/vllm repository by refining the Model Execution API to use keyword arguments for the trtllm_fp4_block_scale_moe function, improving both code readability and reliability in parameter passing. WPC also addressed a deployment issue by correcting the CUDA compatibility library load order within the Docker build, ensuring stable runtime behavior for containerized workloads. These contributions leveraged Python and Dockerfile, drawing on skills in containerization, DevOps, and deep learning. The work demonstrated a thoughtful approach to maintainability and production reliability, focusing on targeted improvements that addressed both API clarity and deployment stability in machine learning environments.